AD-A207 386

UCIRITY CLA5$IFlATION OF TWIS PAOE REPORT DOCUMEIITATI6N PAGEIa. REPORT SECURITY CWSIFICATION lb. RESTRICTIVE MARKINGS 11 L ~J~

2 3. DISTRIBUTIONiAWAJ1.ABITY OF REPORT

ApAroved for public release;23 distribution unlimited.

AD-A207 386 S. MONITORING ORGANiZATION REPORT NUMABERjS)

6a, NAME OF PERFORMING ORGANIZATION 6b OFFICE SYMBOL 7a. NAME OF MONITORING ORGANiZATIONMassachusetts Institute of (If applicable)Technology, Civil EngineerinI CCRE/PACT U. S. Army Research Office

k. ADDRESS (City, State, end ZIP Code) 7b. ADDRESS (City, State, and ZIP Code)77 Massachusetts Avenue, Room 1-175 P. O. Box 12211Cambridge, MA 02139 Research Triangle Park, NC 27709-2211

8*. NAME OF FUNDINGISPONSORING Sb. OFFICE SYMBOL 9. PROCUREMENT INSTRUMENT IDENT,-rATION NUMBERORGANIZATION (If applicable) j 3 & - 7U. S. Army Research Office

8c. ADDRESS (City, State. and ZIP Code) 10 SOURCE OF FUNDING NUMBERSP.0io 21 PROGRAM PROJECT TASK 'N' NOP. 0. Box 12211

Research Triangle Park, NC 27709-2211 ELEMENT NO. NO. NO S 1;

11. 1IT LE (Include Security Classification)

Design Methodology for Automated Construction Machines M A I U I

12. PERSONAL AUTHOR(S)

Slocum. Alxander H.! Demsetz, Laura A.; Levy, David H.; Schena, Bru13a. TYPEOFREPORT 13b.-IE OVERED 114. DATE OF REPORT (YearMothDay) S. PAGE COUNT ....

echnical I FROM 87 TO 1/88 December 11, 1987 65 ....16. SUPPLEMENTARY NOTATION The view, opinions and/or findings contained in this report are those

of.the authr($).and should not as an fficialDeartment of the Army position,17. COSATI CODES 18. SUBJECT TERMS (Continue on reverse if necessary and identify by block number)

FIELD GROUP SUB-GROUP Construction Automation; Construction Machines; Robotics;Machine Design

19 ABSTRACT (Continue on reverse if necessary and identify by block number)

Poor productivity in the construction industry has led

researchers to consider the automation of various construction tasks.

Automated installation of partition wall framework and cement block

walls is addressed here. The wall building task and constraints

imposed on machine design are discussed along with the design of a

pair of machines which automate framework installation.-,,

20. DISTRIBUTION IAVAILABILITY OF ABSTRACT 21. ABSTRACT SECURITY CLASSIFICATION0")UNCLASSIFIED/UNLIMITED 0 SAME AS RPT. 0 DTIC USERS Unclassified

22a. NAME OF RESPONSIBLE INDIVIDUAL 22b. TELEPHONE (Include Area Code) 2;c. OFFICE SYMBOL

DD FORM 1473,84 MAR 83 APR edition may be used until exhausted. SECURITY CLASSIFICATION OF THIS PAGEAll other editions are obsolete. UNCLASSIFIED

UNLSSFE

.UNCLASSIFIEDSI[CURITY ICL.AUIPICATrlON O; TNIS PAGE

(- Preliminary analysis and testing indicate that these machines

provide a cost effective method of constructing partition walls. The

cement block wall building process also lends itself to automation,

and the design of a machine to accomplish this task is also discussed.

The principal common denominator among the majority of

construction automation projects studied was found to be "how to

optimize automated surveying systems to minimize mechanical

complexity".

Accession Fror C

NTIS RA&IDTIC TAB ~~~ -

Unannounced 0iJustification-

0 G

NV NSPEC'r

Distribution/ 6

Availability CoeS

Avafl and/orDiat special

A-L

UNCLASSIFIED

11 SECURITY CLASSIFICATION OF THIS PAGE

Design Methodology for

Automated Construction Machines V

Final Report for the period V

1/1/87-1/1/88

by:

Alexander H. SlocumGeorge Macomber Career DevelopmentAssistant Professor of Civil Engineering

and

Laura A. Demsetz, David H. Levy, Bruce SchenaGraduate Research Assistants

December 11, 1987

U.S. Army Research OfficeResearch Initiatives ProgramContract #DAAL0386GO197

Department of Civil EngineeringMassachusetts Institute of Technology

Cambridge, Massachusetts

Approved for public release;Distribution Unlimited

1_

The view, opinions, and/or findings contained in this report are thoseof the authors and should not be construed as an official Departmentof the Army position, policy, or decision, unless so designated byother documentation.

2

ABSTRACT

Poor productivity in the construction industry has led

researchers to consider the automation of various construction tasks.

Automated installation of partition wall framework and cement block

walls is addressed here. The wall building task and constraints

imposed on machine design are discussed along with the design of a

pair of machines which automate framework installation.

Preliminary analysis and testing indicate that these machines

provide a cost effective method of constructing partition wails. The

cement block wall building process also lends itself to automation,

and the design of a machine to accomplish this task is also discussed.

The principal common denominator among the majority of

construction automation projects studied was found to be "how to

optimize automated surveying systems to minimize mechanical

complexity".

KEY.Wo.QIs Automation, Construction, Design, Machines,

Surveying, Robots

3

TABLE OF CONTENTS

Page1. Introduction 6

1.1 Background 62. Automating installation of interior framework 7

2.1 Wall Installation .72.2 Determining the best degree of automation 11

2.2.1 General considerations 112.2.2 Layout 112.2.3 Track installation 112.2.4 Stud installation 132.2.5 Drywall installation and finishing 14

2.3 Machines to install partition wall framework:The Wallbots 142.3.1 Machine base 142.3.2 The Trackbot 152.3.3 The Studbot 192.3.4 Trackbot and Studbot control 23

2.4 Evaluation 233. Blockbot: A machine for automated construction

of cement block walls 243.1 Wall building methods 25

3.1.1 Economic assessment 283.2 Design constraints and considerations 30

3.2.1 Supplying blocks to the machine 303.2.2 Dimensional metrology system 313.2.3 Manipulator design constraints 313.2.4 State of the art technology assessment 32

3.3 Conceptual designs 333.4 Sizing components for the three principal axes 523.5 Fabrication of a prototype manipulator 583.6 Summary and concluding remarks 58

4. Conclusions 625. Publications and technical reports 636. Participating scientific personnel 64Acknowledgements 64References 64

4

LIST OF FIGURES

PageFigure 2.1. Typical hallway. 8Figure 2.1.1 Cut-away view of wall. 10Figure 2.3.1 Trackbot schematic rear view. 16Figure 2.3.2 Trackbot under construction. 17Figure 2.3.3 Trackbot schematic top view. 18Figure 2.3.4 Studbot schematic rear view. 20Figure 2.3.5 Studbot during preliminary tests. 21Figure 2.3.6 Studbot cam gripper. 22Figure 3.1.1 Application of Surewall Tm surface bonding

compound. 26Figure 3.1.2 Staircase wall construction method. 29Figure 3.3.1 Design #1, 3 Axis Cartesian with 3 DOF wrist. 34Figure 3.3.2 Design #2, 3 Axis Cartesian with 3 DOF wrist. 35Figure 3.3.3 Design #3, 3 Axis Cartesian with 1 DOF wrist.

and 2 DOF base. 36Figure 3.3.4 Design #4, 6 DOF Hybrid with telescoping arm. 37Figure 3.3.5 Design $5, 7 DOR Hybrid with telescoping

arm and travelling base. 38Figure 3.3.6 Design #5, 3 Axis Cartesian with 3 DOF wrist. 39Figure 3.3.7 Design #7, 3 Axis Cartesian with 1 DOF wrist

and 2 DOF base. 40Figure 3.3.8 Design #8, 6 DOF hybrid with 3 DOF wrist. 41Figure 3.3.9 Design #9, 6 DOF hybrid with 1 DOF wrist

and 2 DOF base. 42Figure 3.3.10 Design #10, 6 DOF hybrid with 1 DOF wrist,

2 DOF base and live Z-axis. 43Figure 3.3.11 Design #11, 6 DOF anthropomorphic arm. 44Figure 3.3.12 Lockable universal joint. 48Figure 3.3.13 Schematic side-view of Blockbot. 50Figure 3.3.14 Schematic end-view of Blockbot. 51Figure 3.4.1 Theta arm configuration. 53Figure 3.4.2 Detail of X-Y axis interface. 54Figure 3.4.3 Z-axis servomotor mount. 55Figure 3.4.4 Front view of Z-axis assembly. 56Figure 3.4.5 X-axis cross-section. 57Figure 3.5.1 Completed prototype arm. 59Figure 3.5.2 Control stand. 60

5

1. INTRODUCTIONThis final report describes results of research performed to

develop a design methodology for automating construction processes.Two robots, the Wallbots and Blockbot, were designed and laboratoryprototypes built to help guide the development of the methodology.Based on their development, an assessment is made on commondenominator mechanical, electrical, and sensor systems. One of themost critical technologies identified with respect to successfulimplementation is automating surveying techniques for real timecontrol of construction machinery.

1.1 BACKGROUNDThe U.S. construction industry represents roughly 8% of GNP

[1]. Yet it has notoriously low productivity (40% lower in 1982 thanthe average for private industries [2], and decreasing 1.5% annuallyover the last 10 years [3]). Workers must contend with harsh andoften dangerous conditions. As much as half of their work week maybe spent idle or performing ineffective labor [4].

This combination of low productivity and difficult workingconditions in a large industry has motivated researchers toinvestigate automation of construction tasks. In Japan, where themajor construction firms each have R & D budgets in excess of $10million [3], prototype machines have been developed for shotcreting,fireproofing, concrete finishing, rebar placement, positioning ofstructural members, and tunneling [3,5]. U.S. construction firmstypically allocate little or no funding for R & D; progress to date inconstruction automation has primarily been through researchprojects carried out at educational institutions. At Carnegie MellonUniversity, machines have been developed to aid in the inspectionand cleanup of the Three Mile Island nuclear power plant [6] and toexcavate buried pipes [7]. In addition, four general purpose robotsfor building construction tasks [1] and the automation of sandblastingand concrete formwork cleaning have been considered [8].

A methodology for the efficient automation of constructionprocesses has been proposed by researchers at M.I.T. [9]. Itaddresses the decomposition of construction processes into specifictasks and the evaluation of current technology to determine whethera particular task should be automated or assigned to constructionworkers. Cooperation between architects, machine designers,building contractors, and materials suppliers is emphasized, alongwith the reduction in automated sensing requirements possible

6

through use of information available from previously completedtasks whenever possible and reliance on humans when necessary.

The work described in this report was performed as part of ademonstration of the applicability of this design methodology to atypical building construction process. Characteristics deemedimportant in selecting the process included the following: complexity(the process should be complex enough to divide into tasks, yetreasonable to address in one to two years with limited studentmanpower), repeatability (the process should require fairly regularplacement of materials), independence (automation should bepossible without disrupting the traditional building process), and size(the construction and testing of prototype machines should befeasible on a limited budget in restricted laboratory space).

2. AUTOMATING INSTALLATION OF INTERIOR FRAMEWORKThe installation of non-load bearing partition walls using

standard steel stud framing in buildings of post and slab designsatisfies the above stated requirements for automation susceptibility.Installation of walls is a fairly repeatable process; observe theregularity in the long hallway shown in Figure 2.1. Preliminarydiscussions with contractors on site indicated that the individualtasks involved could be automated without much disruption to thebuilding process. And, compared with most construction materials,the members used in partition walls are fairly lightweight -- a 10foot (3.0 m) section of track or stud weighs under 5 lbs. (22 N), a1/2" x 4' x 8' (1.3 cm x 1.2 m x 2.4 m) sheet of gypsum wallboardweighs 55 lbs (245 N).

2.1 WALL INSTALLATIONThe process of installing partition walls can be divided into the

following five tasks:

Layout--Working from a drawing, the hallway location ismarked, and a line chalked on the floor where the wall is to beplaced.Track installation--Ten foot sections of U-shaped steel trackare placed end to end along both the floor and ceiling. Eachsection is fastened to the floor or ceiling 2" (5 cm) from eachend and intermediately at a spacing not exceeding 24" (61 cm)on center using concrete stub na-ils, shielded screws, cr powerdriven fasteners [10). Installation of the ceiling track is doneby workers on ladders. The track sections must be accurately

7

I

*I i~I

Figure 2.1 Typical Hallway

8

aligned, both along the floor and ceiling and between them, orthe finished wall will be wavy or angled.Stud installation--Steel U-shaped studs positioned verticallyin the track complete the support system for the wall. Precutholes provide passage for wires. The studs are spaced either16" (41 cm) or 24" (61 cm) on center, depending on ceilingheight and layers of drywall used [10]. The studs are fastenedto the track flange using self-tapping screws or by crimping.Studs are fastened either (1) on one side at the ceiling and theother side at the floor, (2) on one side at the ceiling and thesame side at the floor, or (3) on both sides at the ceiling andboth sides at the floor. The third method is required for studsadjacent to door frames and partition intersections [10].Accurate positioning of the studs is necessary to provideattachment locations for the drywall. The on-center spacing ofthe studs should be accurate to 1/8" (3 mm); the verticalalignment to 1/8" (3 mm) over the length of the stud (8' to 12',2.4 m to 3.7 m) [II]. Again, ladder work is required forattachments to the ceiling track.Drywall installation--The gypsum board which forms thewall surface is placed against the frame formed by the trackand stud and fastened using steel screws. Each panel ofgypsum board is generally 4' (1.2 m) wide and the full heightof the hallway. The panels are positioned so that each edge isplaced over a stud, and are fastened to each stud that theycover. Screws along each stud are spaced on either 12" (30 cm)or 16" (41 cm) centers [12]. The gypsum board is heavy, bulky,and fragile. Ladder work is again required.Finishing--After the drywall is fastened, the seams betweenthe panels are filled with taping or all-purpose compound, jointtape is pressed in place, and the excess compound is wipedaway. A second coat of compound is applied and sanded ifnecessary to provide a smooth ever, plane [12].

Figure 2.1.1 shows a cut-away section of wall. Wall constructionusing track, stud, and drywall results in a smooth, sturdy wall.Power and data transmission lines may be installed through thestuds before the drywall is put in place.

An alternative method of construction uses prefabricatedpanels. While a detailed comparison of this method with steel studconstruction may be desirable in the long run, this project's

9

Stud Drywall Sheets

Track Nail affixing

track to floor

Figure 2.1.1 Cut-away view of wall.

10

requirement of minimal disruption of the construction processdictates the use of the current method of construction.

A machine designed to perform any of the wall-building tasksdiscussed above must satisfy constraints on size, weight, anddurability dictated by the environment for which it is intended.Specifically, it must be able to operate on walls from 8' to 10' (2.4 mto 3.0 m) high (only a small percentage of sites have hallways 12'(3.7 m) high), yet fit through a standard doorway (36" wide x 80"high, 0.9 m x 2.0 m). In order to be transported between floors, itmust fit into a standard construction elevator. For several of thebuilding types in which wall-building machines might be used(hospitals, libraries, schools), concentrated floor loads specified bythe Uniform Building Code are as low as 1000 lbs (4.4 kN)[13];machine weight (including materials) should therefore be under1000 lbs (4.4 kN). In addition, any machine intended for applicationon a construction site must be designed to withstand both a dustyenvironment and unintentional loading that may occur duringtransportation and set-up. Finally, any degree of automationproposed must hold reasonable promise of competing economicallywith both union and non-union labor.

2.2 DETERMINING THE BEST DEGREE OF AUTOMATIONThere are several degrees to which the wall-building process

could be automated. General considerations are presented below,followed by a discussion of issues relevant to each task. A survey ofdrywall contractors provided some of the information used here [11.

2.2.1 GENERAL CONSIDERATIONSIn many buildings, room structure is very regular. A large

percentage of walls could therefore be installed by machines capableof only straight line motion. This greatly simplifies machine design.

Obstacles along the path of the wall complicate automatedinstallation. While several of the contractors surveyed noted thatinstallation of HVAC ducts and/or plumbing may occur before layoutand framing installation, others held the opposite to be true. Thesensing required to detect these obstacles, and the flexibilitynecessary to work around them, would increase machine cost andcomplexity. The machines were therefore designed for use whenframing can be installed prior to HVAC ducts and plumbing.

Even when mechanical support systems do not obstruct thepath of the wall, it is unreasonable to assume there will always be a

I1

clear space in which machines can work. A construction site is acomplex, cluttered environment; there will surely be unforseeableobstacles to any machine. Machines will require either elaboratesensing systems or a human operator to prevent their being trippedup by misplaced tools or debris. The presence of a machine operatorprovides additional advantages. An operator can restock materialsas needed and can monitor both machine performance and siteconditions, making minor repairs and adjustments as necessary.Each of the machines discussed below were designed under theassumption that an operator would be available.

2.2.2 LAYOUTLayout requires movibg to a particular location within the

building as indicated on a drawing and then marking the wallposition. In the future, it is possible that buildings may beconstructed incorporating bar-coded elements or metallic gridsembedded in the floor to facilitate automated location sensing;perhaps then it would be feasible for a machine to move itself to aparticular location within a building and lay out a wall. At present,though, the layout task must be left to a human. Locations should,however, be marked in a manner easily detectable by machine.

The method we have chosen uses a laser beam rotating aboutan axis perpendicular to the wall to produce a vertical plane of lightcoinciding with desired wall location. This laser unit is available as astandard surveying tool. A photodetector array produces a currentproportional to the location at which the laser beam strikes it.Mounted in an appropriate location on a machine, it can be used todetect the intended wall location.

2.2.3 TRACK INSTALLATIONTrack is currently supplied to the site in bundles of ten 10 ft.

(3.0 m) long, U-shaped sections with every other section inverted.Two different degrees of automation were considered for trackinstallation. In the first, the channel-shaped track is both formed(from a continuous roll of flat metal) and installed by the machine.In the second, the machine installs (precut) 10 ft. (3.0 m) sections oftrack.

Requiring the track installation machine to form the trackoffers two advantages. First, a minimum space is required formaterial storage, both on the machine and on route to the site. Inaddition, the accuracy with which the machine must be positionedalong the wall's length is reduced, as it is no longer necessary to butt

12

a piece of track up against the previously installed section There aretwo main drawbacks to this approach. Equipment currentlyavailable for roll forming in a fabrication facility is heavy andexpensive (roughly 1000 lbs (4.4 kN) and $10,000 per roll former).While smaller, lighter roll formers could be designed, addressing thisproblem was not reasonable given the time constraints placed on theproject.

Installing precut sections of track is desirable in that currentlysupplied materials are used and heavy roll forming apparatus is notrequired on the machine. The disadvantages of this method are theneed for better control of the machine's position (to ensure that thereis not a large gap between adjacent track sections) and the largerstorage space required for materials.

Since the roll forming equipment is prohibitively heavy, it wasdecided to pursue the second approach and design a machine capableof installing precut sections of track.

2.2.4 STUD INSTALLATIONStuds are currently supplied to the site in the same manner as

tracks, bundles of ten studs with alternate studs inverted. Again,two levels of automation were considered-- manual installationfollowed by automated fastening, and automated installation andfastening.

Dividing the stud installation task between man and machinewas considered because the task split cleanly into types of motions:those easily performed by machine and those more readilyperformed manually. A human can easily pick up a stud and snap itinto place between the flanges of the track. However, ladder work isrequired to attach the stud at the ceiling. A machine can easily bedesigned to reach the ceiling, but design of a mechanism to insert thestuds seemed difficult. Since studs will remain in place temporarilywithout being fastened, it would be possible to have a worker insertstuds between the track flanges and design a machine to align andfasten them. This would eliminate ladder work while simplifyingmachine design.

Although this second approach might be viable, the completeautomation of the process would not significantly increase design ormaterials costs. It was therefore decided to investigate the design ofa machine to both install and fasten studs.

13

2.2.5 DRYWALL INSTALLATION AND FINISHINGAutomated drywall installation presents a major material

handling problem. Sheets of drywall are shipped to the site in astack with good sides facing alternate directions for protection. Thesheets are large (relative to the space available for a machine) andfragile. Material for 50 ft. (15 m) of wall weighs at least 700 lbs. (3.1kN),depending on sheet size, making design of a machine that cancover even this distance without restocking difficult. In whatseemed to be the most feasible configuration for a drywallinstallation machine, sheets of drywall are stored upright at a nearvertical angle with their 'good sides' facing the same direction. Themotions required for the operator to stock this machine are similar tothose required by a worker to manually install the drywall. Sincethe machine would have to be restocked every 50 ft. (15 m), theamount of manual material handling involved would not besignificantly different from what is currently required.

It seemed doubtful that a machine would be economical tooperate under these conditions, and we have decided not to considerthe automated installation of drywall at this time. Since apreliminary study of the feasibility automated joint taping wasinconclusive, we chose to focus our initial efforts on automating theinstallation of the track and stud framework.

2.3 MACHINES TO INSTALL PARTITION WALL FRAMEWORK: THE

The following sections contain descriptions of the two machineswe have designed for the installation of partition wall framework,the Trackbot and the Studbot. Each machine carries an aircompressor on board; electric power is provided through a cablefrom a source on site (survey results indicated distance from poweron site averaged 125 ft. (38 m) [11)). As noted above, layout muststill be done manually. Photodetector arrays mounted on theTrackbot allow it to use the rotating laser for guidance. The laser canbe removed following track installation, as the Studbot relies uponthe installed track for guidance. Floor plans entered into eachmachine's database prior to the start of a job indicate door and studlocations. Eventually, an algorithm for the generation of an optimalinstallation sequence could be incorporated.

2.3.1 MACHINE BASEAlthough each of the machines is designed to perform a

distinctly different function, both the Trackbot and Studbot must be

14

capable of powered motion and steering. Since the machines areapproximately the same size and weight and have similar positioningrequirements, a simple chassis which can be used in both machineswas designed. The chassis is 4 ft. long, 2-1/2 ft. wide and 10 in. high(1.2 m x 76 cm x 25 cm). The frame is supported on four solidrubber wheels; the front wheels are steerable. Distance travelled ismeasured by an encoder attached to the rear wheel. Power isprovided by a DC gearmotor through a chain drive. A removablehandle can be used to tow the machine around the site.

2.3.2 THE TRACKBOTThe Trackbot, shown in Figures 2.3.1 and 2.3.2, positions and

fastens the track on both floor and ceiling. The machine consists oftwo independent track positioning systems, an upper for the ceilingtrack and a lower for the floor track. Track sections are loaded inbundles as received from the manufacturer. The track storage binshold material for 100 ft. (30 m) of wall. The center of the bin islocated 23 in. (58 cm) in front of the vehicle center (see Figure 2.3.3)so vehicle motion along the hallway is required in one direction only.

Each positioning system consists of two 2 degree-of-freedomarms (one at the front of the vehicle, one at the rear) and anorientation mechanism. The arms move horizontally (toward thewall) and vertically. Each supports a vacuum gripper and apneumatic nail gun. The photodetector arrays are mounted justabove the vacuum gripper to provide end-point feedback. Theorientation mechanism consists of a piston actuated gripper to locateeach track section precisely with respect to the vehicle, and an

actuator to flip every other section piece of track 1800 (so that theflanges point away from the surface on which it is to be installed).

To install a track section, the arms first pick the top piece outof the bin and move it to the orientation mechanism. Here, the trackis repositioned and, if necessary, flipped. The positioning arms againpick up the track section, raise/lower it close to the floor/ceiling, andposition it horizontally based on feedback from the photodiodedetectors. The nail guns then fire, firmly anchoring the track sectioninto place. The Trackbot moves forward, making two additionalstops for nailing before placing the next section of track. Since theStudbot relies on the track flange for positioning, doorways must becut out manually after the studs have been installed. However, nailsthat fall within doorways are omitted wherever possible.

15

F 1

Upper vacuum gripper

Upper orientation mechanism

Nail guns removec e x Upper bin

for clarity

Upper arm

Lower vacuum gripperAir compressor

Lower orientationmechanism

Lower bin

Figure 2.3.1 Trackbot schematic rear view

16

14L

Figure 2.3.2 Trackbot under Construction

17

Nail locations: Track sections in bin

Third stop Front nail gun

Front arm

Second stop

First stop

Rear arm

Rear nail gun

Previously installetrack section

Figure 2.3.3. Trackbot Schematic top view

18

Horizontal motion of the arms is accomplished using a steppingmotor and ball screw, permitting precise location of the track basedon the output of the photodiode sensors. The photodiode output isalso used to update vehicle steering. Since all vertical stops thatmust be made by the positioning arms are fixed with respect to thevehicle, vertical motion is accomplished using a pneumatic cylinderwith reed switches indicating stop points. The cylinder also acts as ashock absorber, reducing the magnitude of the nail gun reaction forcetransmitted to the horizontal motion assembly.

2.3.3 THE STUDBOTThe Studbot, shown in Figures 2.3.4 and 2.3.5, can position and

fasten up to 100 studs without reloading. The vehicle path guidancesystem measures distance from the track flange to the vehicle. Thestuds are stored horizontally, again loaded in bundles as receivedfrom the manufacturer. The vehicle moves forward along the trackuntil it reaches a position where a stud is to be installed, as indicatedby the floor plan database. A pair of material handling arms pick thetop stud out of the bin and establish its orientation. If the stud isupside-down, it is delivered to a second arm which flips it overbefore delivering it to the installation arm.

Because of this material manipulation and the asymmetryinherent in a stack of studs, a large amount of misalignment may beexpected prior to delivering the stud to the installation arm.Therefore, this arm must be capable of accepting and realigningmisaligned studs. This problem was solved using a cam gripper (seeFigure 2.3.6) which operates internal to the stud flanges. Two suchmechanisms are mounted on the installation arm. The materialhandling arms settle the stud onto the installation arm, partiallyrealigning it. Then, as the cams rotate out to grip the stud, theycenter the stud on the installation arm.

The installation arm flips the (still horizontal) stud so the webof the channel faces the track, rotates it near vertical, moves itbetween the upper and lower tracks, and rotates the stud to avertical position. The piston which operates perpendicular to thetrack then becomes passive, and hence compliant as the stud istwisted left or right into its final position between the track flanges.

The installation arm has a long beam which runs the length ofthe stud and provides the support necessary to twist the torsionallyflexible studs into the track flanges. The direction of this twist is

19

Bin Piston

FlipperBinFlipper............ vacuumarm-

Inst allat ion

Studs

Aircompres sor.

Fig-ure 2.3.4. Studbot schematic rear view.

20

Figure 2.3.5 Studbot during preliminarytests

21

Pneumatic piston

Stud flanges

\Cam gripper

Figure 2.3.6. Studbot cam gripper.

22

!

determined by the desired final orientation of the stud as detailed inthe floor plan database. Another purpose of this arm's long beam isto support the two crimping tools, one at the ceiling, the other at thefloor. Once the stud is in place, the crimping tools fasten both flangesof the stud to the track (fastening configuration 3 described above).Screwing studs into place is a much more common method ofattachment, but reliable automated screw feeders are prohibitivlypriced (approximately $10,000). Commercially available manualcrimpers were not easily automated, so a new mechanism wasdesigned.

2.3.4 TRACKBOT AND STUDBOT CONTROLFor prototype development, each machine is controlled by a

personal computer equipped with digital and analog 1/0 cards. Oncea suitable control strategy is developed, a machine specific EPROMwill be programmed and incorporated into a hardened controllerwhich will include analog and digital I/O and a communicationsinterface. The communications interface will be used by the machineoperator on site to enter floor plans for the current job.

2.4 EVALUATIONWhile a final evaluation cannot be made at this time, the

following preliminary economic analysis indicates the machines'potential value.

This analysis considers only the savings due to reducedconstruction costs on a per foot basis; additional benefits due toimprove quality and reduced construction time are not included.Labor costs for manual installation of the track and stud frameworkare $1.80 per linear foot (30 cm) for 10' (3 m) high wall with 4" (10cm) studs placed 24" (61 cm) on center [14]. Assumptions used inestimating the cost using the Trackbot and Studbot are:

* The total cost to build the two machines is $40,000; thetwo machines sell for $80,000.

* The combined maintenance cost (parts only) for themachines is $20,000 per year (considered conservative).

* Each machine operates at a speed of 2 ft./min. (61cm/min.); design speed is 10 ft./min. (3 m/min.).

* Each machine operates only 16 hours per week; theremainder of the week is spent in transport and set up

* Each machine requires 40 man-hours per week (foroperation and maintenance), at $20/hour.

23

These assumptions result in a maintenance cost of $0.20 per linearfoot (30 cm), and a labor cost of $0.83 per linear foot (30 cm). Thetotal cost using the machines is thus $1.03 per linear foot (30 cm) ofwall. Operating at two feet per minute (61 cm/min.), 16 hours perweek, the machines and crew of two operators are capable ofinstalling approximately 4 times the length of hallway installed by atwo man crew working a 40 hour week.

Use of the Trackbot and Studbot thus results in a savings of$0.77 per linear foot. If used 16 hours per week, the machines caninstall about 100,000 feet (30 km) of wall per year, resulting in apayback time of approximately 15 months, with savings of $77,000per year over the remainder of the machines' lives. Even with theconservative assumptions used, the resulting rate of return oninvestment is on the order of 37% per year.

The Trackbot and Studbot are nearly complete, with testingand debugging scheduled for Summer 1988.

3. BLOCKBOT: A MACHINE FOR AUTOMATED CONSTRUCTION OF

CEMENT BLOCK WALLS 1

This section discusses development of a robot which is capableof dry-stacking precision concrete blocks. The majority of thischapter involves detailed discussions on how to design a largemachine down to the component level.

Although many other methods exist for building walls ofmoderate strength, the block and mortar wall is the most common.According to the National Concrete Masonry Association (NCMA),over 2.5 billion square feet of wall are built each year at a total costof 8.4 billion dollars. As with virtually all assembly tasks, asignificant percentage of this cost is in direct labor; in this casenearly 60%.2

1 This section is a condensation of a Master of Science Thesis doneby Bruce Schena, "Design Methodology for Large Work VolumeRobotic Manipulators: Theory and Application" Master of Sciencethesis, Department of Mech. Eng. MIT, June 1987. Bruce was one ofProf. Slocum's graduate students.2 Kevin Callahan, Vice-President of Technical Services, NationalConcrete Masonry Association, Herndon, Virginia, October 1986.

24

The construction of concrete block walls is slow, strenuous,repetitive, and a potentially dangerous task for human workers.These characteristics make it an excellent candidate for automation.In order to be economically feasible, however, this or any assemblyoperation must increase productivity or improve the quality of thefinal product. In this case, it is doubtful that any increase in thequality of the finished wall could offset initial capital investment ofan automated system. Therefore, for such an application to be cost-effective, significant increases in productivity are required.

Large and moderately sized companies could potentially save asignificant percentage of labor costs by utilizing automated assemblytechniques to aid in the construction of block walls. These savingscould also result in decreased construction time and improvedquality. It is felt that the most practical application of thistechnology is in building long, relatively low, continuous sections ofwall. Warehouse, factory building, and sound abatement walls alongurban highways are good examples of this type of construction.

3.1 WALL BUI!LDTNG METHODSIt would be very difficult to achieve the productivity levels

necessary to offset the investment in automated block layingmachinery using conventional block-and-mortar constructiontechniques. This is due to the fact that the rate of increase of wallheight per day is not generally a function of just the mason's skilllevel, but rather also of the setting time of the mortar between thecourses (rows) of block. If the maximum number of courses-per-dayis exceeded, the additional weight can cause compression andsubsequent extrusion of the mortar between the lower courses. Thismay result in a wall which is uneven, out-of-plumb, or unstable tothe point of collapse.

Therefore, in order to economically automate block wallconstruction, it is necessary to develop new technologies whicheliminate the physical limit on current construction techniques.Fortunately, two such technologies already exist. The first is afiberglass-reinforced bonding cement called Surewall® which can beapplied to both sides of a block wall in a manner similar to plaster orstucco, as shown in Figure 3.1.1. This fully certified technique,called "surface-bonding", results in a wall with structural propertiesas good, or better than, a conventionally built wall. The mainadvantages of this technique are 1) the blocks can be dry-stackedwithout the use of mortar and 2) the bonding cement can be applied

25

4 *

-.- 7.

-- A

,/ T "4-

A r -r _ laA p M T7_*am

Figue 3..1 Aplictionof Srewal T

surfae boning cmpoun

(From~~ Fin HoeuligCntuto ehius

Tauto Prs, uy94,p8

4.26

to the wall face using a spray gun. Because this technique eliminatesthe use of mortar, the maximum rows-per-day limit is alsoeliminated. In addition, because the surface bonding cement can bespray-applied, it is conceivable that the complete constructionprocess could be automated.

Because this technique has obvious merit in terms ofautomation, it stands to reason that this could also boost productivityin manual wall construction as well. There are two technical reasonsthat this technique is not used more widely. One of these problemsis not relevant when automated machinery is used for constructionand the second can be solved by using a new type of precision blockmanufactured in Sweden.

The first problem is that surface-bonded block construction iscurrently only certified for use in single and two story construction.This eliminates it from use in large building applications. Althoughthis is a problem for conventional construction, it is not forautomated construction. This is due to the fact that currentmechanical technology has practicality limits in the one to two storyheight range already. This is mainly due to structural rigidityproblems associated with a tall machine. Even if the Surewalltechnique could be used for 3 or 4 story construction, roboticplacement of the block at this height would probably be technicallyimpractical. Thus, the two techniques (surface bonding andautomated placement machinery) are ideally compatible from thispoint of view.

The second problem associated with dry-stacking block forsurface bonding is traditionally available blocks are manufacturedaccurate to only ± 1.5 mm (±1/16"). This is mostly due the wear ofthe block molds during manufacturing 3 . The blocks at the beginningof the run are undersized while the ones at the end of the run about300,000 blocks total, are oversized.

This discrepancy between block sizes does not present much ofa problem when using mortar between the courses because theblocks can be individually leveled, but can cause problems for dry-stacking. There are two ways around this problem. The first is using

3 Interview with Doug Buss, Plant Manager of Tarmac Florida,Tarmac manufacturing facility, N.W. 79th Avenue, Miami, Florida,January 1987.

27

some type of shim between the blocks and the second is using specialblocks currently manufactured in Sweden, that are made for dry-stacking. These Swedish blocks, called Leca' m blocks, are preciselyground to exact dimensions following molding. This produces uniformblocks which can be dry-stacked without the problems associatedwith built up tolerances.

For the purpose of this case study, it was assumed that bothsurface-bonding and precision blocks would be used in theconstruction of the robotically assembled wall. Further, it wasassumed that the wall would be built in a "staircase" configurationand does not contain any doors, windows, or thermal expansionjoints. Although this may seem overly simplistic, the addition ofsuch discontinuties would only affect the design of the placementalgorithm and not the mechanical design of the machine.Construction of a long continuous wall (sound abatement walls alongurban freeways are a perfect example) would proceed in the mannerdepicted by Figure 3.1.2. The exact dimensions of the "staircase" willbe determined by the mechanical constraints of the manipulator.

3.1.1 ECONOMIc ASSESSMENTWith the following fixed-value assumptions about the

operating scenario, the required cycle time per block as a function ofthe number of operators, payback period, maintenance percentageper year, and the initial cost of the robot was determined4 .

1) Manual block placement rate of 200 blocks/day2) Total labor cost of $45 per hour, 40 hours/week3) Machine operating 4 hours/day, 4 days/week4) One block moved per cycle

The lowest cycle time was found to be 10.8 seconds. This correspondsto a 6 block-per-minute (BPM) placement rate and the followingoperating conditions:

1) 1 year payback period2) $200,000 initial purchase price of robot3) 50% maintenance cost per year (i.e. $100 K/year)4) 3 human robot operators

4 See Section 2.2, "Analyzing Market Potential"

28

I DIRECTION OF

CONSTRUMTON

6.4"

.. .. ..... ... .... .. .... ...

Figure 3~~...1.2 S.. ras ....... tu tin eto............. .... ....

•~~~~.. ....i I I I

Thus the economics of the application indicate that the robot, underworst-case conditions, must place 6 BPM, 4 hours per day, 4 days perweek. This was thought to be a technically feasible number. In fact,allowing for a margin of error, the target speed for the robotdesigned in this case study was 8 blocks per minute.

3.2 DESIGN CONSTRAINTS AND CONSIDERATIONSIn order to maintain high efficiency, the machine must have a

large enough work volume so that it can keep itself busy betweenmajor changes in the global coordinates. The question of course is"How large is large enough?" Currently available, stock machine toolcomponents such as ball screws 5 and linear bearings quickly excludeany machines larger than about 2 m (6') per axis. Machines largerthan this may require very expensive custom components. Acommon rule of thumb in manipulator design is for every Kg ofpayload, approximately 10 Kg of machinery is required. In this casethe payload is a concrete block (reasonably large by conventionalrobot standards) which weighs 150 N (35 lb). Using the rule ofthumb, the manipulator will weigh approximately 150ON (350 lb).Add to this the weight of any auxiliary equipment such as blockconveyors and the total system weight easily approaches 3800 N(850 lb). This is in addition to a lift needed to move the system tothe proper height and position along the wall.

In order to move this much weight around a construction site arelatively large mobile base is required. Fortunately, mobile, electro-hydraulic scissor lifts are very common on construction sites. Atypical lift has a top platform 3 m long, a payload of 4500 N, and areach height of 8 m. Thus, placing a precision, high-speed robot ontop of one of these lifts seems to be a very attractive solution to thelarge scale positioning requirement.

3.2.1 SUPPLYING BLOCKS TO THE MACHINETwo systems for supplying block to the machine are possible:

The first is an extendable conveyor system with a fixed loadingstation. In this configuration, workers (either human or robotic)would unload blocks from pallets and put them on a conveyor whichwould deliver them to the placement robot. The loading stationwould be in a fixed location for a given section of wall. The secondmethod is to load a pallet of blocks onto the robot vehicle or anaccompanying trailer. This pallet could then be unloaded by the

5 Precision rack and pinion drives can be easily pieced together toachieve almost any length of travel.

30

robot itself, or an assistant (human or robotic) as the machine movesalong the wall.

3.2.2 DIMENSIONAL METROLOGY SYSTEMIn order for the robot to place blocks in the correct position on

the wall, it must know exactly where it is in relation to a knownglobal reference frame. This would be achieved through a large scalelaser metrology system. This system could be based on currentlyavailable rotating laser beacons, large linear diode arrays. ElectronicDistance Measuring Instruments (EDMI), and precise electronictheodolites (angle measuring devices). By implementing a self-checking frame and utilizing redundant measurements, dimensionalclosure of the system can be checked. This will enable precisemeasurement of the position and orientation of the robot in theglobal reference frame.

It takes all these different types of measuring systems, asopposed to a single sensing unit, because no single system currentlyexists. Thus by using several types of systems, measurement closurecan be assured.

3.2.3 MANIPULATOR DESIGN CONSTRAINTSThe number of degrees of freedom (DOF) required for the robot

are directly related to the task that it is trying to accomplish. In thiscase, a full six degrees of freedom are required to place a block on awall. Three of these DOF's are used to translate the load to a positionin space while the other 3 are used to orient it rotationally. For thisapplication the 3 DOF's used for translation must be large, while the 3used for orientation are small since that are only used to compensatefor the pitch, roll, and yaw errors associated with the lift.

The geometry of a typical scissor lift itself imposes a seriousconstraint on the manipulator design. Because the top platform of afully collapsed lift is 1.3 m off of the ground, any manipulator placedon top of the platform must be able to reach down the side in orderto place the lowest course of block onto the slab or footing.

In order to achieve any reasonable accuracy of blockplacement, some sort of end point measurement system is required.Typically, it would be located on the platform that raises the blocklaying head into position. Thus the block laying head should have anaccuracy, even when loaded by its own weight, on the order of twicethat of a typical block (1/16") or 0.03".

31

Besides the static deflections of the structure, there are threeother concerns. The first is the lowest natural frequency of thestructure which should be 3-5 times higher than the operation itperforms. This avoids inducing large oscillations in the structureduring placement of the block. The second concern is the stresslevels in the structural elements. As in any good design, the dynamicand static stress levels should be well within the elastic region of thematerial to insure accurate positioning capabilities and long life.

The final concern is that of thermal effects. Such affects cancontribute to overall structural deflections, large internal stresses,and binding of motion components. In particular, where a machinemay be subjected to environmental extremes, it is important toconsider not only expansion of single components, but also thedifferential expansion of mating components. Large thermalmismatch leads to serious performance problems in sensitiveelements. Thus kinematic design principles should be adhered towhenever possible.

Another very important aspect of manipulator design isconsideration of how it will be manufactured and assembled.Neglecting to think through the assembly process completely canlead to serious problems at a later point in time, usually when theycost a lot of money to fix. Ease of assembly must also be consideredwith respect to maintenance issues. It is all too common for a repairtechnician to be infuriated by a piece of equipment that wasdesigned assuming it would never break down or wear out. Thus,assembli; issues are continuously addressed in this case study.

3.2.4 STATE OF THE ART TECHNOLOGY ASSESSMENTThe complete system to automate the construction of concrete

block walls consists of four major subsystems:

1) The hydraulic scissor lift used for coarse positioning inthe global reference frame.

2) The block supply system such as a conveyor belt orsupply trailer

3) The large scale dimensional metrology system and controlelectronics

4) The large work volume block laying robot

The first item can be purchased directly from a supplier andeasily retrofitted with a more accurate control system. The second

32

item, the conveyor system, is well within the scope of currentlyavailable conveyor technology and thus will not be addressed here.The large scale laser metrology system can be similar to onesdeveloped for machine tool applications. It is the detailedmechanical design of the large work volume robot that will bediscussed in the remainder of this case study.

Although there are hundreds of manipulators on the market,there are few that are suited for the application of building blockwalls. Traditional factory robots are typically designed to operate ina horizontal plane, performing operations such as loading machinetools or palletizing parts. There are very few which are designed toperform the majority of their work in a vertical plane.

Typical revolute-axis robots which may be large enough tohandle a payload of 200 N (one concrete block and gripper) canweigh 4500 N or more and only have a reach of 1-1.5 m. For thesereasons, they are also unsuitable for use on a construction site. Themost practical solution to this problem is thus to design a taskspecific robot which is specifically designed to build block walls.

3.3 CONCEPTUAL DESIGNSA six degree-of-freedom manipulator is required for this task.

However, three of the degrees of freedom only require smallrotational motions: Ox, ey, and Oz. These motions have the following

minimum values because of positioning ability of the hydraulicscissor lift: Ox = ± 2 degrees, ey = ± 5 degrees, Oz = ± 5 degrees. The

Ox and By values are from the design specifications of a typical

construction lift. If the top platform of the lift pitches more than 2degrees or rolls more than 5 degrees, an alarm sounds on the liftindicating unsafe operation. The Oz range was determined throughactual tests on the steerability of a construction lift; it was foundthat the lift could be positioned within ± 5 degrees relative to adesired direction of motion. However, a larger range on Oz would bedesirable for picking a block off the delivery system.

Two broad categories of manipulators for this purpose can bedefined and are shown in Figures 3.3.1 - 3.3.11. The first is the"conventional" manipulator which uses the first three DOFs (countingoutwards from the base) to position the load and the final three insome form of a "wrist" used to orient the load. The second category

33

SIDE VIEW END VIEW

7 z

y /-Y

x

Lift Platform Three Roll Wrist

Advantages: Disadvantages:

Easy to control (cartesian) Complicated and heavy wristLow power consumption Restricts conveyor placementGood load capacity Difficult to sealRelatively rigid (wide bearing spacing) Long arm may vibrateEasy to service Vulnerable to abuse

Hard to keep ways parallelModerately difficult to wire

Figure 3.3.1 Design #1, 3 Axis cartesian with 3 DOF wrist

34

SIDE VIEW END VIEW

z

z

Y


Easy to control (cartesian) Cantilevered arms may vibrateEasy to build Complicated and heavy wristLow power consumption Difficult to sealOffers flexible conveyor placement Moderately difficult to wireRelatively compactEast to service

Figure 3.3.2 Design #2, 3 Axis Cartesian with 3 DOF wrist

35

SIDE VIEW END VIEW

z

Y

2 DOF Base1 DOF Wrist


Easy to control Cantileverd arm may vibrateLow power consumption Not compactLow tip mass Difficult to sealRelatively servicable Moderately difficult to wireOffers flexible conveyor placement Complex baseSimple wrist

Figure 3.3.3 Design #3, 3 Axis Cartesian with 1 DOF Wrist and 2 DOF Base

36

SIDE VIEW END ViEW

5 (Rotation

6 (Rotaton)


Easy to seal Wiring difficultOffers flexible conveyor placement Low stiffnesRelatively compact Complex kinematics

Difficult to build telescoping armDifficult to controlLarge rotational inertiaHigh power consumptionHeavy wristLow load capability

Figure 3.3.4 Design #4, 6 DOF Hybrid with Telescoping Arm

37

I B i ! • II

SIDE VIEW END VIEW

4 (Translation) 2

7 (Rotation)


Easy to seal Difficult to buildOffers flexible conveyor placement Difficult to controlRelatively compact Complex kinematicsImproved stiffness over #4 Arm stiffness may still be a problem

Difficult to build telescoping armHeavy wristLarge rotational inertiaLow load capabilityHigh power consumptionWiring not simple

Figure 3.3.5 Design #5, 7 DOF Hybrid with Telescoping Armand Travelling Base

38

SIDE VIEW END VIEW

z

Three DOF Wrist


Easy to control Complicated and heavy wrist

Easy to build Poor conveyor interface

Low power consumption Large cantilever loads and moments

Easy to service Only moderate stiffnessWiring complicatedDifficult to seal

Figure 3.3.6 Design #6, 3 Axis Cartesian with 3 DOF Wrist

39

SIDE VIEW END VIEW

.z.........

1 DOF Wrist2 DOF Base


Easy to control Not very compactTip mass reduced Poor conveyor interfaceEasy to build Large cantilever loads and momentsLow power consumption Only moderate stiffnessEasy to service Wiring complicated

Sealing difficultComplex base

Figure 3.3.7 Design #7, 3 Axis Cartesian with 1 DOF Wrist and 2DOF Base

40

SIDE VIEW END VIEW

AxA_z

x

Swinging Arm


East to control Complicated and heavy wristGood Conveyor interface High loads on X and Z axesEasy to build High rotational inertiaMedium load capability Stiffness may be a problemServiceability good Sealing moderately difficult

Wiring moderately difficult

Figure 3.3.8 Design #8, 6 DOF Hybrid with 3 DOF Wrist

41

SIDE VIEW END VIEW

@721 zx x

2 DOF Base


Easy to control Large loads on X and Z axesEasy to build High rotational inertiaWrist simpler than #8 Moderate stiffnessGood conveyor interface Difficult to sealServicability good Moderately difficult to wireMedium load capability

Figure 3.3.9 Design #9, 6 DOF Hybrid with 1 DOF Wrist and 2 DOF Base

42

II

SIDE VIEW END VIEW

exx

\o9ez--


Good conveyor interface Difficult to controlLarge loads on X and Y axesRequires precise alingment of

Y-axis slidesLarge e x actuatorLarge inertia on Z axis motorDifficult to wire and serviceComplex design

Figure 3.3.10 Design #10, 6 DOF Hybrid with 1 DOF Wrist, 2 DOF Base

and Live Z-Axis

43

SIDE VIEW END VIEW


Easy to seal Low arm stiffnessOffers flexible conveyor placement Complex kinematicsRelatively compact Difficult to control

Large rotational inertiaHigh power consumptionHeavy wristLow load capability

Figure 3.3.11 Design #11, 6 DOF Anthropomorphic Arm

44

includes the "unconventional" designs. These are manipulators whichhave the first two axes compensate for errors in Ox and ey (i.e. "level

the robot"), followed by three DOFs which translate the load, andfinally a Oz axis which performs the final orientation of the load

about a vertical axis.

Based on a rough analysis of the DOF requirements and workvolume, a number of conceptual designs were generated. Briefly,these are: 6

Design #1: 3 axis Cartesian with a 3 DOF wrist (gantry-type).Design #2: 3 axis Cartesian with a 3 DOF wrist (compact-

type).Design #3: 3 axis Cartesian with a 1 DOF wrist and a 2 DOF

base.Design #4: 6 DOF hybrid with telescoping arm.Design #5: 7 DOF hybrid with telescoping arm and travelling

base.Design #6: 3 axis Cartesian with 3 DOF wrist (cantilevered).Design #7: 3 axis Cartesian with 1 DOF wrist and 2 DOF base

(cantilevered).Design #8: 6 DOF hybrid with 3 DOF wrist.Design #9: 6 DOF hybrid with 1 DOF wrist and 2 DOF base.Design #10: 6 DOF hybrid with 1 DOF wrist, 2 DOF base, and

live Z-axis.Design #11: 6 DOF anthropomorphic.

Each design presented in the figures is annotated with a subjectiveassessment of its advantages and disadvantages. Some of theheuristic rules that were used in evaluating the designs were:

1) Linear axes are more difficult to seal than rotary axes.2) Cartesian manipulators are easier to control than revolute

ones.3) The more complicated the wrist, the harder it is to wire.4) The shorter the cantilever length(s) the stiffer the

structure.5) The shorter the cantilever length(s) the higher the load

capability.

6 Figure #s 15.3.# correspond to design #. Note that the sizes of theelements are only guestimates at this point.

45

6) Gravity back-drivable joints require more power to actuateand need failsafe brakes.

7) Revolute robots are more difficult to service than Cartesianones.

Designs #1 and #10 both have parallel linear bearings mountedon the top platform of the scissor lift. Twisting and bending of thelift could cause accuracy and binding problems in these bearings. Forthis reason, these two designs were eliminated.

Since this machine must be relatively compact and robust inorder to be moved around and used on a construction site, the fixedcantilever arm on designs #2, #3, #6, and #7 is highly impractical.Any contact between this arm and a fixed object (such as acompleted block wall) could cause significant damage to the wall,machine, or a human operator. For this reason (and others) thesefour designs were eliminated.

Design #4 was also eliminated on the grounds of structurallimitations. In this case, these limitations were due to the requiredlength of the telescoping arm. An arm this long (2 m) is likely tohave a very low natural frequency as well as mechanical accuracyproblems. In addition, extremely large actuators would be requiredto move the arm up and down and to rotate the trunk of themanipulator. The pure anthropomorphic configuration, design #11,was also eliminated for the same reasons.

Design #5, although it reduces the problems associated with thelong telescoping arm, is impractical from both an economic andcontrol standpoint. The 7th degree of freedom adds undue expenseand complication to the actuation and control of such a robot. It wasfelt that a more economical solution could be found.

Only 2 designs remain at this point, #8 and #9. These two arein fact very similar; the only major difference being how the 3rotational axes are implemented. While #9 would have problemswith a distorting top lift platform, design #8 would have to carryaround a very heavy, bulky wrist in order to move the turret.Neither of these solutions is particularly attractive. A reasonableway around these problems is a hybrid combination of the two:Remove the turret, replace it with a single block gripper, andeliminate the powered Ox and ey motions. If these motions are

46

replaced by a passive universal joint-type mechanism between thegripper and the powered Oz axis, then the block will automaticallyorient itself with gravity if it is held at its center of mass. Since thisis what the two servo axes would be doing anyway, it seems like areasonable solution.

Replacing two servo axes with passive joints has obviousadvantages. It also has some potential disadvantages. Threeforeseeable problems include:

1) Error due to variation in block center of gravity.2) Error due to bearing friction.3) Swinging of the block during block transfer from conveyor

to wall.

It is felt that all of these problems can be reduced or eliminated byusing a specially designed joint and precision blocks.

If precision blocks are used (as previously assumed) thevariation in the location of the center of gravity should be negligiblysmall. Since the mix used to produce these blocks is extremelyhomogeneous, large variations in material density from one side ofthe block to the other are not anticipated. Block to block variationsin total weight are unimportant.

Assuming the following about bearing friction:

* 0.001 coefficient of friction for quality roller bearings* 12.5 mm diameter universal joint shaft* 150 N block* 50 N gripper• 200 mm distance from block CG to joint axis

the total lateral positioning error for the block, due to bearingfriction, is approximately 0.025 mm. This is acceptable for thisapplication.

Swinging of the block during the pick and place operation canbe prevented with the use of a locking universal joint sketched inFigure 3.3.12. The universal joint is mounted in the center of ahollow-rod, double-acting pneumatic cylinder. When the cylinder isactuated in one direction, three fingers lock the joint in a position

47

Rotary Actuator

Hollow RodPneumaticCylinder

Spherical Seat

Motion - Solid Hemispherical

During DomeLocking,

Gripper Mount

Clamping Fingers

Universal Joint

Figure 3.3.12 Lockable Universal Joint

48

perpendicular to the Theta arm. In this position, a block can bepicked up from the conveyor and servoed to position along the wallwithout swinging. When it is near its final position on the wall, thecylinder is released, allowing the block to swing free and align itselfwith gravity. If desired, the cylinder could be then actuated in thedownward direction which would clamp the hemispherical dome inits current position. Final servoing and block placement could thenoccur. Following the release of the block, the joint would be rigidizedagain and the process repeated.

It is felt that this design would have the following benefitsover a 3 axis powered wrist:

* Reduce manipulator tip weight and inertia* Eliminates 2 expensive servo-actuators and coupling

hardware0 Eliminates 2 servo systems and electronic hardware* Reduces system power consumption

This somewhat unconventional configuration thus seems to be thebest design.

Figures 3.3.13 and 3.3.14 schematically show the finalconfiguration for the robot axes. In general, the manipulator has 6DOF; four of which are powered. Gross horizontal and verticalmotions are provided by the X and Z linear axes. Motions in and outfrom the wall are provided by the swinging "Theta arm". The fourthand last powered DOF is the 8 z motion at the wrist of the

manipulator. The two unpowered degrees of freedom are generatedby a hanging universal joint connected to the Oz motor. This overall

configuration has the following attractive features:

* Easy to build* Easy to control* Easy conveyor interface* Reasonable to wire• Reasonable to service* Compact to transport* Uses gravity to orient blocks

49

* f .. 4

C40)

500

4-

00

4--b)

51)

S 6

3.4 SIZING COMPONENTS FOR THE THREE PRINCIPAL AXESBefore machine components could be sized, the acceleration

profiles were calculated for the various axes to achieve the targetspeed of 8 blocks per minute.

X-axis: 2 meters in 3 secondsZ-axis: 0.3 meters in 1 secondShoulder: 180 degrees in 1.0 seconds (negotiable)Wrist: 90 degrees in 0.6 seconds (negotiable)

Under static operating conditions, the joints of a roboticmanipulator can be assumed to be servoed to the desired positionsby the control system. 7 This assumption allows the joints to beregarded as rigid and any deflection at the end-point due todeflection of the structural elements alone.

The most severe loading condition for the robot occurs whenthe Z-axis is accelerated upwards (+Z) while the Theta arm is at fullextension and carrying a concrete block. With an estimated tip massof 36 kg (including the arm structure and load), an appliedacceleration of 1.4 m/s 2 (derived from the cycle times of section 5.3),plus the acceleration due to gravity, the worst-case tip load is about400 Newtons. With a reasonable safety factor of 1.5, the design load,applied at the tip of the structure, is 600 N (150 lb). This is thevalue used to calculate the Z-direction deflection of the entirestructure. The deflections of the structure in the Y and X directionsare not critical since both can be easily servoed with respect to avertical plane of laser used to indicate straightness of the wall. 8

The detailed engineering analysis done during the course of thedesign of the prototype is too extensive to present in this paper, butis available in the form of a Master of Science Thesis 9. For reference,Figures 3.4.1 through 3.4.5 show the mechanical detail of theprincipal assemblies.

7 This of course is assuming that the position control loop has anintegrator and a reasonably high gain.8 Hint for designing the dimensional metrology system.9 Bruce M. Schena, "Design Methodology for Large Work VolumeRobotic Manipulators: Theory and Application" Master of Sciencethesis, Department of Mech. Eng. MIT, June 1987.

52

Shoulder BracketZ

Wrist Actuator Shoulder Actuator

Locking Cylinder(in s e c tio n ) ------ ------ -----

2 DOF Universal Ball Slides

Joint

Block

Gripper

Figure 3.4. 1: Theta Arm Configuration

53

Ezb-.

4)

U

Cu 0

Cu

4) d)

en0

czV

54

zTiming Belt Drive

X

Ball Screw Support

Gusset (hidden)

Ball Screw Z-Axis Motor

Motor Bracket

Z-Axis Box Beam V

Figure 3.4.3: Z-Axis Servomotor Mount

55

Upper baliscrew support Actuator bracketcarriage

Linear guidemounting pad

Ball screw

Y

Linear guide

;0

8"

Lower ball screwsupport (Thin section bearing)

Bellows seal

Cantilevered bellows section (u wy

Figure 3.4.4 Front view of Z-axis assembly

56

Bellows stiffening plate Bellows Seal

(in section)

Bellows guide rail

Z-Axis Mounting

I z! % B a l l s c r e w

i , Linear guide

lO"x4"x3/16" Steel box beam

Figure 3.4.5 X-axis cross-section

57

% % 0 0

3.5 FABRICATION OF A PROTOTYPE MANIPULATORIn order to demonstrate the concept of stacking concrete blocks

with an automated system, a prototype manipulator was built.Photographs of the finished arm and control stand are shown inFigures 3.5.1 and 3.5.2 respectively.

Due to financial, time, and laboratory space constraints, onlythe shoulder and wrist axes were built to specification. A simplepneumatically-actuated Z-axis was also built so that blocks could bepicked up. Specifically, the prototype had the following features:

0 Full-size/full-capacity pneumatic gripper• Full-size theta arm, exactly as designed* (1) DJDE-02, 96:1 actuator with resolver* (1) DJDE-04, 196:1 actuator with resolver* Full-size DJDE-04 mounting bracket* Pneumatically-powered Z-axis (20" stroke)* 5" x 5" x3/8" steel box tube test stand• (2) Moog 152 Series PWM controllers* (1) Moog 150 Series Power Supply* (1) Moog 3 winding power transformer* (1) IBM PC-AT for real-time control* (1) 96 channel digital 1/0 BaseBoard* (1) 6 channel A/D and 2 channel D/A Data Translation

Board

Intentionally left out of the prototype were:

0 Full-size X-axis* Full-size Z-axis* Locking wrist universal joint* Scissor lift platform

All arm components were machined by the author using in-house facilities. All parts were machined from the materials asspecified in this thesis. Following completion of the hardware, a real-time control system was implemented.

3.6 SUMMARY AND CONCLUDING REMARKSThe final configuration for a robot which is capable of

performing concrete block wall building task was selected from afield of 11 conceptual designs. This final design consisted of 2 linearaxes, 2 rotary axes, and 2 unpowered rotary wrist axes. The exact

58

Figure 3.5.1 Completed Prototype Arm

59

Figure 3.5.2 Control Stand

60

range of motion of each axis was determined by available ball screw,linear guide, and scissor lift technology. The structural elementswere sized through static deflection and stress analysis usingCastigliano's energy methods while the lowest natural mode of thestructure was calculated using a lumped-parameter model. Thisnatural mode was found to be dominated by the torsional stiffness ofthe shoulder actuator.

The linear guides and ball screws for the linear axes wereinitially sized using standard heuristic rules and manufacturer's data.The sizing of these components was then independe.Lly checked bythe manufacturer's applications engineer. It was found that althoughthe target design speed could only be achieved using custom ballscrews, it would be much more economical to build a prototype usingstock screws.

Dolan-Jenner Corp.'s precision rotary actuators were chosen forthe two rotary axes of this robot. These drives were sized using theprocedure recommended by the manufacturer. The two linear axismotors were sized by calculating the load power requirements andcomparing this to the output power of a variety of motors. Twobrushless DC Moog motors were selected to drive these axes.

Based on the experience of building two of the axes of thisrobot, a preliminary cost estimate of an entire robotic blockplacement system was made. This included the following directequipment costs:

Motors (4 axes) $10,000Control hardware (4 axes) 10,000Control computer 5,000Structural components 10,000Linear guides 3,000Ball screws 5,000Misc. mechanical 10,000Misc. electrical 10,000Laser metrology system components 20,000Construction scissor lift 25,000Block supply system (trailer type) 15,000Bonding cement spraying equipment 15.000

Total materials costfor quantities of one item: $138,000

61

The target (retail) price was $200,000. Whether the device could bemanufactured in quantity for less than $100,000 would require amore detailed economic analysis by marketing and manufacturingexperts that is beyond the scope of this work.

4. CONCLUSIONSDesigning the Wallbot and Blockbot construction robots has

facilitated the identification of the following common parameters:

Mechanical systemsThere is a need for lightweight kinematically designed linear

servoactuator-bearing assemblies that can be assembled with othercomponents in module form. Most sub-assembly components (e.g.bearings, ballscrews, and motors) exist as individual stock items, butthere should be modular interface units that allow the easierassembly of these components into a design. This will facilitateeconomical design and fabrication of dedicated modular constructionrobots that would be easy to repair under harsh field conditions.

Electrical systemsElectrical system components are already available in the

modular "plug-in" mode that is desired for mechanical components.Technology in this area is thus entirely adequate.

Sensor systemsConventional surveying systems are far too slow, inaccurate, or

require too much human input to make them useful as sensorfeedback elements for real time control of construction robots. Todate, the only surveying tool that is useful for constructionautomation is the rotating laser beacon and light mast that is oftenused to control blade height during precision grading operations.Other sensors such as tilt meters, theodolites and EDMs are either tooslow, too inaccurate, or require too much human effort to beeffective. The attainment of a fully active Global Positioning Systemwill help layout location of system boundries, but will not be fastenough to achieve the 0.001 second update times required tofacilitate real time control of robots. Thus research is needed in thearea of fast, automated, accurate sensors for angle and distancemeasurement with part per million accuracies.

Overall Conclusions

The principal conclusions to this work are thus:

62

1) Construction robots can be designed to effectively increaseproductivity and quality of specific construction tasks.

2) General purpose construction robots (e.g. robots whichemulate a human worker) are neither technologically or economicallyfeasible, and will not be for the forseeable future.

3) Because only dedicated machines are technologically oreconomically feasible, the technology for designing dedicatedconstruction robots is well known by experienced designers ofmachine tools and robots.

4) The principal research endeavors needed to advance thestate-of-the-art of construction robots are:

a) Identification of which processes to automate. This requiresa thorough detailed analysis of the construction process on a case bycase basis for all types of construction processes. Only then can thego-ahead for the design of a specific robot be given.

b) Development of advanced angle and distance measuringsensors with order of magnitude greater accuracy and speed than arepresantly available. Because sensors are so expensive, the numberrequired needs to be minimized in reduce cost, but moreimportantly, to reduce the chances of damage on-site. Maximizingthe autonomy of any given construction robot requires an increase insensor accuracy so the robot can perform for a longer period before asensor is needed to pick up where another sensor's accuracy is leftoff. Research in this area could overlap with the Strategic DefenseInitiative program.

5. PUBLICATIONS AND TECHNICAL REPORTS1) Currently writing Art and Science of Precision

Mechatronic Design, Prentice Hall Pub. Co. EnglewoodCliffs N.J. to appear in Fall 1988.

4) Slocum, A.H., Demsetz, L., Levy, D. Schena, B. andZiegler, A., "Construction Automation Research at theMassachusetts Institute of Technology", presented at thethird Int. Symp.sium on Construction Robotics, June 22 -

24, 1987, Haifa Israel.3) Schena, Bruce, "Design Methodology for Large Work

Volume Robotic Manipulators: Theory and Application",S.M.M.E, Sept. 1987.

4) Levy, David, "Studbot: A Construction Robot for theAutomated Assembly of Steel-Stud Partition Walls",S.M.M.E, Sept. 1987.

5) Slocum, A.H., Demsetz, L. and Levy, D.. "AutomatedConstruction of Partitioned Wall Framework",

63

6. PARTICIPATING SCIENTIFIC PERSONNEL

Prof. Alexander H. Slocum, George Macomber CareerDevelopment Assistant Professor of Civil Engineering

Ms. Laura Demsetz, Research Assistant, Department of CivilEngineering

Mr. David Levy, Research Assistant, Department of Civil

Engineering

ACKNOWLEDGEMENTS

We wish to thank the following for their generous support: NSKCorporation, Anaheim Automation, Pneutek Inc., and SMC, Inc. Thiswork has been supported in part by the M.I.T. Sloan Basic ResearchFund, the George Macomber Career Development Chair, the LincolnInstitute for Land Policy, M.I.T.'s Project Athena, and principally byaward #DAAL0386G0197 by the Army Research Office under theResearch Initiatives Program.

REFERECS1. Warszawski, A. and Sangrey, D.A., "Robotics in Building

Construction," Journal of Construction Engineering andManagement, Vol. 111, No. 3, September, 1985, pp 260-280.

2. Warszawski, A., "Robotics in Building Construction," Carnegie-Mellon University, Pittsburgh, Pa., 1984.

3. Podolsky, D.M., "Japan Leads U.S. in Construction Research andRobotics," Robotics World, March, 1986, pp 8-20.

4. Whittaker, W.L., "Construction Robotics: A Perspective,"Proceedings of the Joint International Conference on CAD andRobotics in Architecture and Construction, June, 1986, pp 105-112.

5. Hasegawa, Y., "Robotization of Reinforced Concrete BuildingConstruction in Japan," Proceedings of the Joint InternationalConference on CAD and Robotics in Architecture and Construction,June, 1986, pp 113-121.

6. Peterson, I., "Stepping into Danger," Science News, Vol. 130, No. 2,July 12, 1986, pp 28-31.

7. Whittaker, W.L., "Cognitive Robots for Construction," CarnegieMellon University Robotics Institute Annual Report, 1985.

64

8. Skibniewski, M.J., "Engineering and Economic Analysis of RoboticsApplication Potential in Selected Construction Operations," Ph.DDissertation, Carnegie Mellon University, 1986.

9. Slocum, A.H., "Development of the Integrated ConstructionAutomation Design Methodology," presented at the SMEConference on Robotic Systems Education and Training, August11-13, 1986, Detroit, MI.

10. ASTM Standard C754-82.11. Poulin, K., unpublished survey of building contractors.12. ASTM Standard C840-79.13. Uniform Building Code, 1985 Edition, Table 23-A.14. Mahoney, W.D., ed., Open Shop Building Construction Cost

Data,1986, R.S. Means Company, Inc., Kingston, MA, 1985.

65

Documents

AD-A207 386